alt Hacker NewsI thought LLMs were a great tool for learning new topics - perhaps even complex ones. But overtime, I have had several frustrations with this. First, I get exhausted reading LLM prose. I really don't want to read anything generated by something like Opus 5 at this point. Second, as I dive deeper, I need a way to organize the information in a useful way as I begin to branch out in many different directions. I have tried to use the LLM to fix this by having it generate a web page with diagrams and organized information flow. It's an improvement, but I still run into the issues I described in my first pint - LLM prose is annoyingly dense, and the useful information gets lost in a bunch of noise. You can direct it do something like "use plain English and avoid LLM prose - provide only as much information as necessary to demonstrate the point", but it is once again only a marginal improvement.
And then I begin to think to myself that I should just read a book on the topic written by a trusted source who put a lot of effort into teaching the topic properly and presenting the information in a thoughtful way. So, I am back to books and mostly try to use LLMs to clarify certain questions or ideas I have.
Replies
I have had very similar experience! I wanted to learn Probablistic ML, checked out a couple of MOOCs, but didn't find any that were at my level - some were too advanced, some too beginner level. Claude was unable to one-shot a course, so I am now asking it to generate it module by module. But even here, it is not doing a very good job. I muddle through the concepts that it has written, do a whole bunch of back-and-forth, which tbh is exhausting, and then rewrite everything in my words so it actually makes sense to another human being.
> I get exhausted reading LLM prose
So much this! If I see one more sentence with the words "genuinely" juxtaposed with "load bearing" my head is going to explode!
btw, I am building the tutorial here for anybody interested in this topic: https://github.com/avilay/learn-probml
Good to know other people who get migrane reading LLMs dense prose. I started reading books again recently, since everything online is polluted by LLM prose. What i realise, is that a human author, especially a teacher understands the learning pathways of new learners, they motivate the learning, and start from simplest concepts (a spherical cow), and then building all the complexities. This helps us to emphasize on most important concepts, while throwing away unnecessary complexities. While reading LLM prose is like reading a research article, that is written to an expert in the area, that talks about bleeding edge, with full of jargons, caveats, that just is not conducive to the learning process for a new learner.
I never used office hours as a student, which I later regretted because it made me work longer and harder to perhaps achieve somewhat better understanding in some classes, but also I dropped every proof-based math course I ever took. Overall I think my education would have been stronger by attending office hours.
I view LLMs in education similarly to office hours. Some people abuse it to get homework answers without grappling with the material, but the optimal amount is not zero.
LLM certainly not a replacement for a book, where you get someone’s extended personal approach to a topic, thoughtfully organized, reviewed and edited, often times actual courses taught based on it, with answers checked and errata available online.
>I really don't want to read anything generated by something like Opus 5 at this point.
Personally, I find that its generated prose tends to have an undue weight to it, almost as if every topic I ask about somehow bears a heavy burden, or is otherwise load-bearing, to use its parlance.
Quite puzzling, really.
I have the exact same experience, so reassuring to know I'm not the only person who feels this way.
I will say, opus 5 is an egregiously bad case of this, but other LLMs have this too, just less bad.
I've decided that the main thing I'm building is my own mental model. You can take notes, create docs, put graphs and websites together, but unless I'm just trying to generate some reference material the only real objective is to develop the understanding and intuitions inside my own brain.
So I have the LLM offer a very short explanation of something, and from there's it's just me asking questions. Anything that feels fuzzy or not fully internalized is something I poke at until I'm satisfied.
It really has helped me develop a sensitivity to what I understand vs what I don't, and the ability to drill into any part of it is amazing.
The sycophancy is also a concern, it’s not really an impartial teacher, all its training is to suck up and maximize engagement rather than learning. The incentives are wrong.
You can make something like the link below. I'd say it's a more than marginal improvement. It's still tiring, but it's much better according to my taste. I imagine everyone would have their own version of this for their own preferences.
It's a loop that uses adversarial review to check several dimensions of the writing:
https://github.com/Vibecodelicious/llm-conductor/blob/main/w...
I use the LLM to point me at relevant books and papers. But there is still a trust problem: I am trusting the LLM to point me at reliable, trustworthy sources.
I'm not sure I'm better off with humans though -- I'm not qualified to judge whether a source is a proper authority, not an I qualified to judge whether someone knows enough to point me to a reliable source.
It seems this is a fundamental epistemological problem to which there may never be an answer.
I have been using LLMs to help me turn my journals into interconnected notes and sometimes it is so confusing to read the notes that it doesn't resemble any human would write. Its like the models are getting stronger while also losing its touch to write human sounding sentences on complex topics.
This feels like the weirdest complaint to me. Just tell it to shorten its response.
Literally saying "one sentence response" solves most of this problem.
I've found it helpfull to ask the LLM to generate a sylybus for the topic. treat the sylybus as a design doc for a price of software. i find they do much better when they have subtasks to focus on. they can do big picture and small picture, but they can't do both at the same time.
Same. I usually just read the book along and ask questions on a specific part, rather than trying to get the LLM to produce an entire study guide for me. It seems to work better as a Q&A than a "teach me" advisor
Even if you tell the LLM not to use LLM pros they (still) do it. If you feed the Wikipedia article on signs of AI writing and tell them to use none of those signs they will also (still) do it. I have tried (many times) to get an LLM to explain a concept to me, or a process, or an algorithm or what have you, and every time they cannot help themselves. Either they use LLM pros, or they get so verbose that it all just becomes noise and I spend more time filtering out unnecessary jargon than I do reading let alone learning anything.
Have you tried using the caveman skill ? :D Might be worth a try if you dont like long prose https://github.com/JuliusBrussee/caveman
> First, I get exhausted reading LLM prose. I really don't want to read anything generated by something like Opus 5 at this point.
I’ve found the tone of Kimi K3 to be less obnoxious. Unfortunately it doesn’t wholly solve the issue, I don’t think any LLMs out there have a truly pleasant writing style, but at least not every assumption is “load bearing”.
Yeah, I am currently trying to work with Opus 5 to refresh myself on deep learning fundamentals, and... it's a mixed bag. I'm glad I already am familiar with the subject matter, as I can prompt for refinement and improvement. It is kinda following the Karpathy videos so far (a couple lessons in) but adding more math/derivations, which was what I asked for. It has trouble staying on topic, presenting information in a coherent/meaningful order, and providing all the context necessary to move through steps in its "course notes".
Like I said, I'm essentially continually prompting to refine the material. LLMs certainly continue to append, and never cut back. It just keeps spitting out additional content at me. So that's a bit annoying too. But I can basically get figure out what's going on with a few extra promps.
If youre curious what i've got so far... just be warned it is quite literally AI slop plus me continually prompting for clarification/cleanup etc. : https://github.com/cmoscardi/ai-for-ai
To get rid of the llm prose issue you can take a representative sample of its prose (say a question and its answer), rewrite the answer in the way you'd prefer, and add that to the system prompt; I did this to get my LLMs to compress down what they say, and now everything they say is very dense and to the point
So you want it to teach you and take notes for you?
I have been using this tool for the past few months that was posted on here: https://github.com/devenjarvis/lathe
It generates tutorials for you, and serves a webpage that lets you complete them. It does a remarkable job.
It still has a bit of the LLM prose problem, but it does help you fine tune the ‘voice’ it uses.
anybody adds some prologue about what style of answers are prefered ? i know i often try to change the linguistic patterns because i too (unsurprisingly) am tired of llm prose.
> want to read anything generated by something like Opus 5 at this point
I've stopped using CC because of it. I find it insufferable.
>LLM prose is annoyingly dense, and the useful information gets lost in a bunch of noise.
This problem doesn't get talked about enough and is second only to the hallucination problem IMO.
AI produces so much noise to wade through in order to find signal, and the more expertise you have in a field the more that costs. That noise directly subtracts signifcantly from productivity gains.
And, I think the problem is directly related to the hallucination problem. It feels very much like an effort to kitchen sink the response in order to provide some value among possible hallucinations.
It also seems to be a byproduct of Gen AI operation. It just fundamentally doesn't understand what it's outputting, so doesn't know how to narrow down to the most salient bits.
"hey LLM, I'm trying to learn about ___. I already know ___. My favorite authors are ___. Suggest some reading material."
It annoys me that the default AI mode is so tedious and longwinded. I read a lot of nonfiction - the house style of AI is basically marketing copy.
something to try that works really well is to put "explain this as if you're talking to a 5th grader" at the end of your request.... it just breaks down the text into more manageable sentences that can be understood by general audiences
Its aligned in getting you interested and frustrated about a topic enough to go to a primary source you would have never looked at
You can instruct it to be terse and even take on a specific voice if the standard prose bothers you.
LLMs help me refine my search. If I want to dive deeper into any topic, I can yield a strong list of primary sources relatively quickly.
> LLM prose is annoyingly dense
It's just long. It just doesn't shut up. It's overly verbose. And you can't tell it to be concise or you degrade its quality.
If I ask what an integral is, the correct answer is that it is the continuos analog of a sum, generally used to calculate areas and volumes.
It should really be a single sentence, and then let me ask more about the terms I don't understand, and here's the beauty, in the previous one there can be only 5 terms I cannot know.
An LLM will vomit an entire page or more of explanation which isn't bad per se, but is an answer to something different: "give me a short introductory explanation to integrals". And that's not what I asked.
I think AI is surprisingly good at this. I use voice mode while working out to learn complex topics, follow up with reading, and then go back to ask the LLM more questions. They excel at simplifying complex ideas and have endless patience. One hack I found is telling the LLM to test my knowledge by asking me questions—that gives me a clear idea of what to read next. Overall, they’re a great tool to use alongside traditional learning methods like reading books and working through practice problems.
AI hater here. Too me this story is all to predictable. This is exactly how I thought using LLMs to learn would go. Or in simpler terms: well duh.
I'm actually going to make a prediction here as well. I think you will soon realize that using LLMs to clarify certain questions or ideas you have will turn out to have frustrations as well. And that you will soon direct those questions to either peers you know in real life or internet forums which are very likely to have a non-AI policy.
[dead]
[dead]
This is on top of the issue of LLMs being a moron. like, I'm sure dumb people can learn stuff from them, but I'm sticking to books written by people who know what they're talking ahout and not vibed slop.
> First, I get exhausted reading LLM prose.
Often "be concise, to the point." is enough, but you can also paste it some stuff you like as an example text and ask to do style transfer.
You can change the prose that it outputs to anything you like. ask the llm how.
It's much better to feed the book to the LLM and ask questions as you read along, instead of asking the LLM to basically write a custom book for you from scratch.